Hierarchical mixtures of experts and the EM algorithm
Neural Computation
Bidirectional Clustering of MLP Weights for Finding Nominally Conditioned Polynomials
ICANN '09 Proceedings of the 19th International Conference on Artificial Neural Networks: Part II
| Hi-index | 0.00 |
This paper proposes a method for finding a set of regression rules to fit data containing nominal variables as well as numerical ones. Here a regression rule is a linear regression function accompanied with the corresponding nominal condition. A set of such rules can be learned by a four-layer perceptron. A couple of model parameters are selected based on the BIC. In our experiments using 11 real data sets, the method exhibits better performance than other methods for many data sets, and found its own significance of existence in the field of regression.